Opportunities and challenges for thermally driven hydrogen production using reverse electrodialysis system

(1)

Opportunities and challenges for thermally driven hydrogen production using reverse electrodialysis system

Yash Dharmendra Raka, Ha˚vard Karoliussen, Kristian Myklebust Lien, Odne Stokke Burheim

^*

Dep. of Energy and Process Engineering, Faculty of Engineering, Norwegian University of Science and Technology, NTNU, Norway

h i g h l i g h t s

First study of levelized cost of hydrogen by ammonium bicarbonate reverse electrodialysis.

For economic feasibility, max. membrane cost 20V/m²and min. membrane life 7 years.

Waste heat required varies from 480 to 105 kWh/kgH2.

InPresent scenariomembrane replacement cost dominates with 44% of total expenses.

InFuture scenariocost of heating dictates with 58% of total expenses.

a r t i c l e i n f o

Article history:

Received 21 December 2018 Received in revised form 11 May 2019

Accepted 13 May 2019 Available online xxx Keywords:

Reverse electrodialysis (RED) Ammonium bicarbonate (AmB) Sustainable energy

Hydrogen production

Low grade waste heat to hydrogen Levelized cost of hydrogen (LCH)

a b s t r a c t

Ongoing and emerging renewable energy technologies mainly produce electric energy and intermittent power. As the energy economy relies on banking energy, there is a rising need for chemically stored energy. We propose heat driven reverse electrodialysis (RED) technology with ammonium bicarbonate (AmB) as salt for producing hydrogen. The study provides the authors’perspective on the commercial feasibility of AmB RED for low grade waste heat (333 Ke413 K) to electricity conversion system. This is to our best of knowledge the only existing study to evaluate levelized cost of energy of a RED system for hydrogen production. The economic assessment includes a parametric study, and a scenario analysis of AmB RED system for hydrogen production. The impact of various parameters including membrane cost, membrane lifetime, cost of heating, inter-membrane distance and residence time are studied. The results from the economic study suggests, RED system with membrane cost less than 2.86V/m², membrane life more than 7 years and a production rate of 1.19 mol/m²/h or more are necessary for RED to be economically competitive with the current renewable technologies for hydrogen production. Further, salt solubility, residence time and inter-membrane distance were found to have impact on levelized cost of hydrogen, LCH. In the present state, use of ammonium bicarbonate in RED system for hydrogen production is uneconomical. This may be attributed to high membrane cost, low (0.72 mol/m²/h) hydrogen production rate and large (1,281,436 m²) membrane area requirements. There are three scenarios presented thepresent scenario,market scenarioand future scenario. From the scenario analysis, it is clear that membrane cost and membrane life inpresent scenariocontrols the levelized cost of hydrogen. Inmarket scenarioandfuture scenario the hydrogen production rate (which depends on membrane properties, inter-

*Corresponding author.

E-mail address:[email protected](O.S. Burheim).

Available online atwww.sciencedirect.com

ScienceDirect

journal hom epa ge: www.elsev ier.com/locate/he

https://doi.org/10.1016/j.ijhydene.2019.05.126

0360-3199/©2019 The Author(s). Published by Elsevier Ltd on behalf of Hydrogen Energy Publications LLC. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

(2)

membrane distance etc.), the cost of regeneration system and the cost of heating controls the levelized cost of hydrogen. For a thermally driven RED system to be economically feasible, the membrane cost not more than 20V/m²; hydrogen production rate of 3.7 mol/

m²/h or higher and cost of heating not more than 0.03V/kWh for low grade waste heat to hydrogen production.

licenses/by/4.0/).

Introduction

Many large-scale industries for e.g. manufacturing applications dissipate heat at relatively low temperature. In general, this heat is in overabundance, and often cannot be reintegrated entirely on-site or used for district heating. Thus, this heat is rejected to the surroundings [1]. Globally, waste heat in the industrial sector accounts for more than 40% of total energy use, meaning almost half of the energy consumed is wasted as heat to the environment[2]. In Norway, the waste heat potential was estimated to be 19 TWh/a in 2008[3]. There are various sources of industrial waste heat as shown in the Table 1.

Though there is a huge potential for waste heat recovery, the amount of heat varies with temperature as shown inFig. 1.

The low-grade waste heat is an untapped energy resource generated in industrial plants due to lack of efficient and cost- effective recovery methods. About 300 thousand USD savings was accounted for carbon dioxide equivalent (CO2,eq) emissions from waste heat from industries in UK[5]. In order to mitigate climate change, EU policy 2030 recommended a target to reduce GHG emissions by 40% and improve energy efficiency by 27% in the transportation and industrial sectors.

Further, the use of hydrogen in the transportation sector was identified to be an alternative solution. We propose an energy efficient ammonium bicarbonate reverse electrodialysis (AmB-RED) system that utilizes the low grade waste heat (333 Ke413 K) to produce hydrogen. Reverse electrodialysis (RED) stack is an electrochemical device that converts chemical energy into electrical energy using concentration gradient across an ion selective membrane. RED is one of the few renewable technologies which is capable of directly producing hydrogen from waste heat.

Cost and some technicalities are major barriers to market penetration of RED technology[6]. Literature survey indicates lack of studies evaluating challenges in market diffusion. The

economic studies report that membrane accounts for almost 40e80% of total investment cost[7e10]. The literature study till date to the best of the authors’knowledge shows quite an uncertainty in estimated membranes costs ranging from 2V/ m²to 640V/m²as shown inTable 2. Moreover, the membrane life was assumed to be in the range of 7e20 years. While some are optimistic assumptions and rest suppositions, an engi- neered estimate of membrane cost for levelized cost of energy (LCOE) for RED needs to be addressed.

In this article, we evaluate for the first time the economic feasibility of AmB RED system for hydrogen production with low grade waste heat of (313 Ke413 K) as a source. A model with technical and economic aspects was developed to evaluate levelized cost of hydrogen (LCH). The model can be applied to decentralized hydrogen refuelling station. Critical parameters were identified, and their sensitivity was evaluated. We estimate the membrane price for the technology to be economically viable for hydrogen production. In this parametric study we account for five parameters. In the scenario analysis, we propose three scenarios estimating the values of the financial and performance indicators for present state, economically competitive to other clean energy technologies and an optimistic future scenario.

System description

A typical RED cell converts directly the electrochemical potential of salt into electrical energy. The electrochemical potential in the form of concentration gradient across an ion exchange membrane provides the driving force for ions to migrate from the high concentrate channel (HC) to the low concentrate channel (LC). By using alternate permselective membranes, the flow of ions is controlled. This ionic current is converted into electrical current at the electrodes by redox reactions. The ammonium bicarbonate based RED system is divided into a RED stack and a regeneration system as shown inFig. 2. Similar to a typical reverse electrodialysis system, a concentrate and a dilute solution of ammonium bicarbonate are introduced into the cell in the respective channels. The concentration difference across a cation exchange membrane causes ammonium ions to flow from HC to LC. Similarly, bicarbonate ions flow from the HC to the LC across anion exchange membrane. Mixing of these solutions decreases the concentration of the HC effluent, while the LC increases. This LC effluent is introduced into the regeneration system. The regeneration system includes a stripping column and an absorption column. The low grade waste heat, at 313 K and Table 1eIndustrial waste heat potential (TWh/a) from

different sectors at various temperature range (their share [4]).

Sector L.T<373 K M.T 373e572 K H.T>572 K

World 63% 16% 21%

Transportation 7.9 (46%) e 9.3 (54%)

Industry 3.7 (42%) 1.8 (20%) 12.1 (38%)

Electricity 26.2 (88%) 3.6 (12%) e

Residential 3.0 (36%) 5.4 (64%) e

Commercial 2.3 (59%) 0.2 (5%) 1.4 (36%)

(3)

Fig. 1eIndustrial waste heat [TWh/a] in Norway as function of temperature [K].

Table 2eSummary of previous economic studies on RED for electricity production.

Author Year Capacity MW Lifetime years CmemV/m² tmemyears PopW/m² LCOEV/kWh

Turek[11] 2007 e e 100^a,b 10 0.46 6.79^a

Post[9] 2010 200 40 2 7 2 0.08

Daniilidis[10] 2014 200 25 50 7 2.2 0.72

Weiner[7] 2015 e 20 750^a,b e 1.2 6.33^a

Bevacqua[12] 2017 100 20 50 20 4.78 0.3

a Currency in $.

bCost includes membranes, gasket, spacers, electrodes, end plate.

Fig. 2eSchematic of Reverse Electro-Dialysis (RED) system based on ammonium bicarbonate salt with regeneration system for hydrogen production (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article).

(4)

1 atm, heats the LC effluent solution in the stripping column.

This heat strips the solution into ammonia and carbon dioxide gases. The stripped gases absorb into the HC effluent in the absorption column at 298 K. The outlets of absorption and stripping column are reintroduced into the RED cell by mixing them with HC and LC inlets.

A typical RED stack consists of K4Fe(CN)6/K3Fe(CN)6 as electrode rinse solution (ERS) with inert electrodes such as Ti coated in IreRueO[13]. The alternative cost efficient carbon electrode with FeCl2/FeCl3as ERS are not used widely due to low conductivity[13,14]. In AmB-RED system, we propose use of 1 M KOH as ERS with Nickel/nickel mesh as electrode for hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) similar to alkaline water electrolyzers, respectively.

The redox reactions at the respective electrodes are as follows.

Reactions at Anode:

H2OðlÞ#2H^þ_ðaqÞþ1

2O2;ðgÞþ2e Reactions at Cathode:

2H^þ_ðaqÞþ2e#H2;ðgÞ

Ammonium bicarbonate has low decomposition temperature of around 60 C at 1 atm that makes it an ideal candidate for low grade waste heat recovery [12,15e18]. Thermolytic salts such as ammonium bicarbonate can access control over solution concentrations. Increase in the salt concentrations increases the osmotic pressure differences which results in higher energy production. An ammonium bicarbonate based RED system can produce osmotic pressure heads equivalent to 380e510 m for higher concentrations (1e1.5 M), which is three to four times higher when compared to seawater-riverwater RED. In addition, use of ammonium bicarbonate salt in RED system avoids the geographic constraints such as coastal re- gions that are essential for sea/river based RED systems. This eliminates the need for energy intensive pretreatment processes. Further, high solubility in water, and relatively low molecular weight, makes it suitable for RED application.

Ammonium bicarbonate production in terms of environ- mental concern, CO2,eqemissions, production of AmB can be net carbon negative [19]. Moreover, carbon capture can be realized in the oxyfuel process, where oxygen and steam are burnt with natural gas. Recovering the waste heat from such processes, not only as hydrogen but additional oxygen too makes AmB RED to hydrogen a carbon neutral economy. AmB has an additional advantage compared to NaCl solutions because its use reduces electrode overpotentials for both hydrogen evolution reaction (HER) and oxygen reduction reaction (ORR)[20]. Thus, AmB may make these reactions more favourable for energy production than previously achieved using river water and seawater in RED systems.

Methodology

Input data

The economic assessment was performed for a decentralized hydrogen production plant (refuelling station) of capacity 1500 kg/day[21]. Depending on the conversion efficiency, the

waste heat required for such capacity varies from 1 MW to 8 MW. List of input parameters for techno-economic modelling are shown in theTable 3. The dimensions of RED stack are 1 m1 m x 2 m (l x w x h) per 10 kW (electric power)[10]. The technical parameters including model equations for membrane resistance and theoretical waste heat for ammonium bicarbonate decomposition are referred from literature [12,22]. It was difficult to extract market price of membranes, due to commercial sensitivity. Hence, the cost estimates were assessed based on the information from the literature (see Table 4).

Assumptions

In this method, inflation rate was moderate and similar to the interest rate. All economic assumptions were based on values taken from literature for European market. In the economic analysis, fuel, insurance and emissions cost, were considered.

Ion exchange membranes (IEM) for RED specific applications follows the learning curve and economic trend as Nafion. The levelized cost of hydrogen (LCH) of 3.59V/kg was considered for market scenario [21]. The operational cost associated to heating including equipment cost heat exchanger was assumed to be 0.01V/kWh[23].

Table 3eInput technical and economic data for proposed model.

Parameter Symbols Value Unit

Concentrate solution conc. chc 2.0 M

Dilute solution conc. clc 0.06 M

Ideal gas constant R 8.314 J/K/mol

Temperature T 293 K

Faraday's constant F 96,485 C/mol

Permselectivity of CEM/AEM a 0.85

Diffusivity coefficient (mem) DAmB 2,10^e12 m²=s Inter-membrane distance d_ch 100 mm

Membrane thickness d_mem 125 mm

Number of membrane pair N 10

Residence time tres 1 s

Length of channel l 0.1 m

Width of channel w 0.1 m

Viscosity of water m 9,10^e4 Pa-s

Porosity ε 0.8

Velocity of dilute sol. vlc 0.01 m/s

Membrane life time tmem 4 yr

Plant life time t 20 yr

Operational hours per year ta 8000 h

Pump efficiency h_pump 0.75

Faradaic efficiency h_F 0.95

Production capacity m^capH2 1500 kg/day

Pumping system cost cpump 300 V/kW

Membrane cost Cmem 150 V/m²

Waste heat price cheat 0.01 V/kWh

Discount factor r 0.06

Infrastructure price cinfra 1123 V/m^2a Regeneration system price cregen 3400 V/m³=h^b

Labour cost Clab 20% (% of CAPEX)

RED stack CREDstack same

as Cmem

V/m²

a Land area.

b Vol. flow.

(5)

System model

The system model was subdivided into AmB-RED system, pumping system and regeneration system. In AmB-RED system, electrochemical equations were used to estimate hydrogen produced. Pumping system model deals with hydrodynamics in the cell and related losses.

Regeneration system was modelled based on simple mass balances and thermal duty assumed from Bevacqua et al.

[12].

AmB-RED system

The objective was to estimate the hydrogen production, and membrane area required. At first, the input data as shown in Table 3 were initialized. Parameters such as conductivity and activity coefficients were estimated. On this basis, cell voltage, resistances of membranes and channels were estimated followed by stack resistance, and stack voltage.

Using these values, short circuit current density, peak power current density, and peak power stack voltage were determined. Finally, the moles of hydrogen produced and membrane area required were estimated. The proposed system was mathematically modelled using equations as follows,

Conductivity of solutions (k): The equivalent conductivity of ammonium bicarbonate depends on concentration and molar conductance which depends on temperature. The conductivity was calculated at constant temperature 293 K using Jones-Dole equation,

k¼L,ci (1)

whereL¼L0^A_1þB^L^,Lⁱ^,c¹⁼²ⁱ

L,c¹⁼²_i CL,ci. HereAL;BL;CLare model parameters used for fitting and referred from Bevacqua et al.[12], ciis inlet concentration;L0molar equivalent conductivity of salt at infinite dilution. The conductivity is calculated in mS/

cm for both feed solutions at 293 K.

Activity coefficient of solutions ðgÞ: The activity coefficients depend on molar salt concentration. The linear dependence of activity coefficients on concentration was estimated using ENTRL-RK thermodynamic package in Aspen Plus.

g¼g1$ciþg2 (2)

hereciis expressed in mol/l, g1¼0.1366 l/mol and g2¼1.0007.

These values were estimated at T¼293 K[12].

Unit cell open circuit potential (E^ocp_u:c:): It is the membrane potential [V] for a cell pair (CEM, AEM) with no losses considered. The electrochemical potential difference across an IEM placed between two different concentration solutions can be described using modified Nernst equation, E^ocp_u:c:¼ ðacemþaaemÞ,R,T

F ,lng_hc,c_hc g_lc,clc

(3) ais permselectivity of IEMs measured at concentrationchcand clcat constant temperature 293 K for specific membrane. Here we assume sameafor both membranes.Fis Faraday constant.

Tis room temperature.Ris ideal gas constant.

Area specific membrane resistance (R_aem=cem): It is the ohmic resistance [Um²] of the membranes when immersed in the solution. It is expressed as function of concentration[12].

Rcem¼Raem¼r1,ðclcÞ^0:236 (4)

Rcem;Raem are area specific membrane resistances for cation and anion exchange membrane, clcis inlet concentration of low concentrate solution.r1is fitting parameter [Um²/M]. It is estimated to be 0.0002 [Um²/M] forpresentandmarketscenario and 0.00004 [Um²/M] forfuturescenario[12]. As concentration lowers below 0.1 M, the contribution to resistance from low concentrate solution dominates.

Channel ohmic resistance (R_lc=hc): The resistance (Um²) due to the solution in the channel and spacer geometry. It depends on concentration, and is calculated using molar conductivity of the salt.

Rlc¼ dlc

ε,klc,clc

Rhc¼ dhc

ε,khc,chc

(5) heredchis intermembrane distance inmm,εis the porosity of spacers,klcandkhcare conductivities of feed solutions

Unit cell resistance (R_u:c:): The cumulative sum of resistance of membranes and channels in a unit cell [Um²].

Ru:c:¼RcemþRaemþRlcþRhc (6)

Current density at peak power (j^pp_u:c:): The peak power current density occurs at potential half the open circuit potential. As the resistance remains constant the current density [A/m²] at peak power is calculated using Ohm's law,

j^pp_u:c:¼ E^ocp_u:c:

2,Rcell

(7) Table 4eEffect of raw material cost and production volume on cost of membrane [28].

Membrane Production rate m²/a Craw$/m² C_prod$/m² Total cost $/m² Thicknessmm

Nafion 2800 e e 576 183

250,000 35 100 e 183

28,000,000 e e 62 183

10,000 e e 250 25

600,000,000 e e 2.8 25

SPEEK 250,000 4.5 50 e 100

NFM 10,000 1þ0.05 10 e 200þ25

(6)

Actual unit cell potential (E^act_u:c:): The potential across the RED unit cell drops due to ohmic resistances in the RED unit cell and is estimated as,

E^act_u:c:¼E^ocp_u:c:Ru:c:,j^pp_u:c: (8) Number of unit cells (N_u:c:): Alkaline electrolyzers operate above reversible potential 1.23 V. The anodic and cathodic overpotentials are estimated to be 300 mV at low current densities [24]. Hence, the required potential for water electrolysis is assumed to be 1.5 V. In order to meet this potential, unit cells are stacked together in series. The minimum number of unit cells to be stacked in series is calculated as,

Nu:c:¼1:5

E^act_u:c: (9)

Stack open circuit potential (E^act_stack): The stack open circuit potential [V] drops upon connecting the RED device to external load. Thus the actual voltage across the terminals of the stack is given by,

E^act_stack¼N,E^act_u:c: (10)

RED stack resistance (R_stack): The total resistance including unit cells, N and electrodes in a stack.

Rstack¼N,R_u:c:þRele (11)

hereRele[Um²] is resistance from electrodes in electrode rinse solution. This resistance value is assumed to be 0.01Um²[12].

But, its contribution to the performance is not significant for large stacks.

Hydrogen production rate (n__H₂): The theoretical moles of hydrogen produced per unit time in the compartment with electrode-electrolyte rinse solution of RED stack.

_

nH2¼j^pp_u:c:,3600

z,F ,h_F (12)

herez¼2 is the ion valence per mole of hydrogen gas,h_Fis Faradaic efficiency. h_F signifies that all current density generated is not utilized to make hydrogen, due to system related losses. In a RED system, the main source of loss in Faradaic efficiency is due to ionic short circuiting in feed and drain channels. This loss in Faradaic efficiency is 6% for well designed RED system in comparison with alkaline water electrolyzer where these losses range from 5 to 25%[20,25].

Membrane area (A^tot_mem): The total membrane area [m²] required to produce 1500 kg hydrogen a day (m_^cap_H

2 ¼0.0174 kg/s is the required production capacity).

A^tot_mem¼I^pp_para:,Nu:c:

j^pp_u:c: (13)

I^pp_para:is total current from the stacks in parallel. It is calculated as I^pp_para: ¼ ^m^_

cap H2,1000,z,F

2:016,hF . Where, m_^cap_H

2 is required hydrogen production capacity [kg/s] and molecular weight of hydrogen is 2.016 g/mol.

Pumping system

Experimentally, a pressure drop is observed over the inlet and outlet of feed water. Hence, to compensate this loss, pumping of feed solutions is essential. The important parameters such as flow rate, pressure drop and pumping power were calculated and are presented in this section.

Flow rate (Q_lc): As suggested in the literature[22,26], the velocity of LC solution influences the double layer resistance and thus the power density. The flow rate [m³/s] for low concentrate feed solution is estimated as,

Qlc¼N,vlc,dch,w,ε (14)

hereεis porosity of the spacer,vlcvelocity of feed solution in low concentrate channel [m/s], d_lc intermembrane distance [m],wwidth of channel.

Hydraulic diameter (d_h;lc): The spacer filaments obstruct the flow through the channel and this requires additional pumping power. To estimate the influence of spacer filaments, hydraulic diameter of spacer filled channel was determined using following equation,

d_h;lc¼ 4ε

2 d_chþ

ð1εÞ,d⁸_ch

(15)

Pressure drop (Dp_lc): Assuming an ideal case of fully developed laminar flow, the pressure drop (Pa) was estimated using Darcy-Weisbach equation,

Dplc¼ 12,m,l² 0:25,d²_h,tres

(16) themis dynamic viscosity of water [Pa-s],tresis the residence time [s],llength of channel.

Pumping power (P_pump): The power [W/m²] required to overcome the hydraulic resistance in pumping the feed solutions through the channels. This strongly depends on the spacer porosity, as it dictates hydraulic radius which influences pressure drop.

Ppump¼Dp,d_h;lc,ε

tres þDp,d_h;hc,ε tres

(17)

Reynolds number (Re): It is the ratio of inertial forces to viscous forces within a fluid. The Reynolds number for wide channel corrected for spacer porosity is defined as [27],

Re¼r,vlc,dh

m (18)

Mass balances

The concentration difference between the adjacent channels drives the salt flux from high concentrate to low concentrate channel.

(7)

Salt flux (J_salt): The total solute transport.

Jsalt¼j^pp

F þ2,DAmB

cⁱⁿ_hccⁱⁿ_lc dmem

(19) here dmem is the thickness of IEMs in operation [m], DAmB

diffusivity coefficient of ammonium bicarbonate in membrane [m²/s]. The first term relates to coulombic (counter ion) transport, while the second term relates to co-ion transport.

The factor 2 relates to number of membranes in a unit cell.

Waste heat/Regeneration system

The regeneration system compensates the amount of salt diffused through the IEM from high concentrate channel to low concentrate channel. The regeneration system includes air stripping column and absorption column. The air stripping column decomposes solution from dilute compartment outlet to ammonia gas and carbon dioxide gas at 333 K. The absorption column dissolves the decomposed gases at 298 K in the outlet of concentrate channel.

Heat required for regeneration (q_regen): The total amount of thermal power required to stripQ_lcof ammonium bicarbonate salt from LC solution

qregen¼q__th,Qlc,3600,A^tot_mem (20)

qthis the specific thermal duty [kWh/m³] required to decom- pose ammonium bicarbonate solution into its components NH4,(g) and CO2,(g). The value was estimated using relation qth¼a1,eâ²^,C¹a3,Câ₂⁴þa5,Câ₁⁶,Câ₂⁷from Bevacqua et al.[12]for inlet concentration to stripping column C1 and outlet con- centrationC2from stripping column. Herea1toa7are fitting parameters that are function ofC1, refer toappendix Table 8.

Economic model

Levelized cost of hydrogen (LCH) method assesses the economic feasibility of the proposed process. This method con- siders annuity factor _ð1þrÞ¹ t

!

as shown in equation(21). This method is preferred as it is simple, and provides realistic results. These results of economic assessment of different technologies are easy to compare, transparent, and easy to understand. The disadvantage of using this method is that the distinction in variations in costs and benefits from one year to the next is not possible (same net benefit is applied to every year). Further, the time delay between investment and first year of regular operation are not considered.

LCH¼ Pi¼t

i¼1CAPEXþOPEXþCheat ð1þrÞ^t þC_mem;re P_i¼t

i¼1 mH2 ð1þrÞ^t

(21)

Capital cost (CAPEX)

The capital cost [V] of the plant was evaluated as CAPEX, using equation(22).

CAPEX¼CmemþCREDstackþCregenþCpumpþClabþCinfra (22) The capital costs include all the expenses related to pur- chase and installation (labour and infrastructure) of systems such as RED stack (including membranes), pumping unit,

regeneration unit. The labour cost includes civil work associated with site preparation, process equipment building, offsite services and is assumed to be 20% of capital cost. The cost of regeneration system was normalized to flow rate [m³/h], and cost of pumping system was normalized to pumping power [kW]. The cost of membranes was normalized per unit membrane area. The cost of membrane was estimated as sum of the production cost and raw material cost. Using analytical approach to investigate the cost of Nafion membranes, Minke et al.[28]suggested that for constant thickness with increase in production rate by four orders of magnitude the total membrane cost decreases by almost one order of magnitude w.r.t the initial cost. With a decrease in thickness by almost 7 times and a further increase in membrane production rate by one order of magnitude, the total membrane cost reduces to one tenth of the present raw material cost[28]. Assuming a learning rate similar to that of Nafion membrane of 25mm thickness, a minimum membrane cost was estimated. The production rate was assumed 6010⁸m²as a RED application requires large membrane areas due to low membrane potential or power density[10].

Cost of RED stack (C_REDstack): It includes cost of electrodes, gaskets, spacers and end plates. It is assumed to be the same cost as that of the membrane.

Infrastructure cost (C_infra): It is the real estate cost and this was calculated using dimensions of RED cell, stack referred from Daniilidis et al.[10].

Pumping system cost (Cpump): The pumping unit consists of two pumps. The cost associated with a pump is cpump

[V/kW]. The pumping unit compensates the pressure drop inside a RED stack, the total cost is normalized per unit stack.

Cpump¼Ppump,A^tot_mem,cpump

1000 (23)

Regeneration system cost (Cregen): The specific investment cost for regeneration system is c_regen [V/m³/h]. The total cost of the regeneration system is estimated for the required flow rateQ_lcfrom a RED stack.

Cregen¼cregen,Qlc,A^tot_mem,3600

2 (24)

Operational and maintenance cost (OPEX)

OPEX [V] includes labour maintenance, service and repairs expenses which is assumed to be 2% of CAPEX. Direct and indirect operational costs such as chemical, emissions to air, supply of water, labour, taxes, administration, and insurance are not included.

OPEX¼0:02,CAPEX (25)

heretis plant life time [years].

Membrane replacement cost (Cmem, re)

The ion exchange membranes have limited lifetime. Hence, these membranes have to be replaced at the end of their lifetime until plant lifetime. The cost associated to membrane replacement is calculated as,

(8)

C_mem;re¼X^n¼f

n¼1

Cmem

ð1þrÞ^t^memⁿ (26) wherecmem andtmem are membrane costs [V/m²] and membrane lifetime [years].fis the factor that denotes the number of times replacement takes place (f ¼t=tmem).

Waste heat price (Cheat)

The regeneration system requires heat for producing NH3,g

and CO2,gfrom the LC outlet solution. This heat required by the process can be sourced from various thermal energy intensive industries including cement, paper, biomass, waste incineration etc. Though it is said to be waste heat, a heat exchange system is required to provide the heat at specific temperature. Hence a costc_heatis associated with it. This cost is assumed to be as low as 0.01V/kWh.

Cheat¼cheat,qregen,ta (27)

tais the total number of operational hours per year [h].

Profit

The sum of revenue earned by selling produced hydrogen and operational expenses incurred on annual basis.

Profit¼ X^i¼t

i¼1

mH2,cH2

ð1þrÞ^t

! X^i¼t

i¼1

CAPEXþOPEXþCheat

ð1þrÞ^t þCmem;re

!

(28) wherecH2is hydrogen selling price [V/kgH2] and assumed to be 3.59V/kgH2[21].

Scenario case study

The study includes three scenarios based on the cost of membrane, the membrane life time and the hydrogen production rate. In all the scenarios for ammonium bicarbonate solution, the hydrogen production rate was calculated from the corresponding input data as shown in Table 5. In the present scenario, the cost and lifetime of membrane were fixed input and LCH was estimated. In themarketcompeting scenario, the LCH was assumed the same as estimated by US DOE for hydrogen produced from renewable energy source.

In themarketandfuture scenario, the concentration of feed solutions were optimized to maximize hydrogen production for present properties of membranes refer Table 5 [12].

There are no profits made inmarketscenario, the membrane life was assumed to be 7 years and membrane cost was estimated. Finally, an optimisticfuture scenariowas considered, wherein the profits of 10% were made on annual basis, with an increased membrane life of 10 years, and the membrane cost was estimated accordingly. In the future scenario, the hydrogen is produced at 3.71 mol/m²/h with increased permselectivity of membranes 0.95 and Faradaic efficiency 0.95, as shown inTable 5. The channel length and width were assumed to be the same for all scenarios. The heating cost for all scenarios was assumed to be 0.01V/kWh.

The Table 6 summarizes all the input and estimated parameters.

Results and discussion

Cost of membrane

Fig. 3a shows the influence of membrane cost on the LCH for AmB based RED system for hydrogen production at 0.72 mol/

m²/h production rate, and a membrane life,tmem of 4 years.

LCH is a linear function of membrane cost,Cmem. The LCH of AmB RED system decreases with decrease in membrane cost, C_mem.Fig. 3b compares LCH for ammonium bicarbonate RED system in the marketand the future scenarios. The rate of decrease of LCH formarket scenariois three times as high as future scenario, this shows that LCH was sensitive to membrane cost inmarket scenario. The intercepts inFig. 3b shows the contribution of balance of plant cost to LCH at membrane cost,C_mem¼0V/m², and, LCH at this intercepts are below the LCH target set by US DOE. Assessing all the costs inpresent scenario, it was found that the membrane cost influences the most, accounting for almost 40% of CAPEX. This is in agree- ment with the previous studies [9e11,29]. The minimum predicted cost of membrane was calculated using the learning rate as that of 25mm Nafion membrane for production rate of 6010⁸m²/a. The minimum membrane cost was found to be 1.69 V/m². For ammonium bicarbonate RED system, the membrane cost,Cmem, must drop below 2.86 and 22.3V/m²for membrane lifetimet_memof 4 and 10 years at production rate of 1.19 and 3.71 mol/m²/h, respectively, to compete with other clean energy technologies.

Hydrogen production rate

The LCH is predicted for different hydrogen production rates.

With increase in production rate, the LCH decreases. With the present membrane cost (150 V/m²) and membrane life of 4 years, the LCH decreased by almost 42% (72.49V/kg) for increase in production rate from 0.72 mol/m²/h to 1.19 mol/m²/h.

This increase in production was due to optimal concentration Table 5eInput parameters of AmB RED system for scenario study.

Parameter present market future Unit

Concentrate solution conc. (chc) 2 2.6 2.6 M Dilute solution conc. (clc) 0.06 0.05 0.07 M

Permselectivity (a) 0.753 0.85 0.95 e

Inter-membrane distance (d_ch) 270 100 100 mm

Residence time (tres) 70 60 50 s

Membrane lifetime (tmem) 4 7 10 years

Faradaic efficiency (h_F) 0.95 0.95 0.99 e

Table 6eList of input and output parameters for scenario study.

Scenario Input Output

present Cmem,tmem n_H2, LCH, Profit market LCH,tmem, Profit n_H2,Cmem

future Profit,tmem n_H2,Cmem, LCH

(9)

of feed solutions. The LCH decreased to 23.7V/kg for further increase in hydrogen production rate to 3.71 mol/m²/h. This increase was a combined result of increase in permselectivity and Faradaic efficiency. The increase in production rate increases the total hydrogen produced in the plant lifetime.

Therefore, LCH decreases. Increase in production rate is due to increase in current density, which indirectly relates to concentration ratio. High concentration ratios are achieved with high solubility of salt under the assumption that the electrical conductivity of the solution and permselectivity of membrane remains same. The production rate is a function of peak current density, which in turn depends on Nernst potential. As this Nernst potential relates to ratios of concentration or activity coefficients which are salt properties (solubility etc). The salts that can produce hydrogen at a production rate higher than that of AmB-RED system due to synergetic effect of higher OCP, higher solubility limit and lower resistance needs to be inves- tigated experimentally. However, this is beyond the scope of this work. Hydrogen production rate can be increased by improving membrane properties such as permselectivity, conductance and reducing thickness while maintaining the mechanical strength. Further, the use of optimal combination of electrolytic solutions, electrode material and their geometric configuration can decrease cell overpotential and increase the overall stack potential, which in turn will improve the hydrogen production rate.

Lifetime of membrane

Fig. 4a describes the influence of membrane lifetime at a membrane cost of 150V/m²forpresent scenario. With increase in membrane life, the LCH decreases substantially. As the membrane lifetime increases from 3 years to 10 years, the LCH decreases to almost 40% of the initial value in the present

scenario.Fig. 4b shows asymptotic behaviour oft_memrequired at a hydrogen production rate of 1.19 mol/m²/h to 3.71 mol/

m²/h in order to meet a LCH of 3.59 V/kg. The increase in membrane life 7 years or more, show less significant impact on LCH in presentandmarket scenario. Due to the operating conditions, the membrane properties such as permselectivity, membrane resistance etc. can be expected to deteriorate with time. As the performance of the system depends on the membrane properties, corresponding deterioration in the system performance will be observed. The higher the deteriorate rate, the quicker the membranes need to be replaced. For best performance, the system is operated at optimum conditions, and to maintain the performance, the membranes need to be replaced on timely basis. Increasing the frequency of replacing a membrane increases membrane replacement cost (contributes 44% of total expenses inpresent scenario). Thus, the overall trend implies that the lifetime of membrane (indirectly, the degradation rate) of these membranes plays a critical role in minimizing LCH. market scenario becomes economically feasible at membrane life of 7 years while 1 year forfuture scenario. This shows that increase in membrane life beyond 7 years has relatively low impact on LCH with the case parameters chosen in the study.

Inter-membrane distance

Increase in the inter-membrane distance decreases the actual unit cell potential. The decrease in open circuit potential decreases the peak power current density, which in turn decreases the salt flux. Hence, the theoretical heat required also decreases. On the contrary, the increase in inter-membrane distance cause, an increase in feed solution volume flowing through the channel. This, again, causes an increase in the theoretical heat required to regenerate the ammonium Fig. 3ea. Influence of cost of membrane on LCH for a membrane lifetime of 4 years in case ofpresent scenario(left). b.

Comparison ofmarket,futurescenario with US DOE target price, for LCH as function of membrane cost (right).

(10)

bicarbonate solutions. This increase in the heat requirement nullifies the decrease in heat requirement due to decrease in salt flux. Thus, the net effect observed is an increase in the total heat required.

Moreover, as the increase in inter-membrane distance increases the feed solution flow rate. The increase in flow rate increases the amount of water per gram of salt. Thus, the total heat required to recover same amount of salt increases.

Increased inter-membrane distance has a negative impact on LCH, as shown inFig. 5a and b The LCH decreases with decrease in inter-membrane distance. The increase in inter- membrane distance increases the channel resistance. This increase in resistance decreases the open circuit potential and increases number of membrane pairs required. This results in increase LCH. In case of themarket scenario, the LCH varies almost linearly with change in inter-membrane distance. In themarket scenarioand thefuture scenariothe maximum inter- membrane distance to match with the US DOE were found to be 100mm, and 200mm fromFig. 5b respectively.

Cost of heating

Fig. 6a and b, shows the influence of cost of heating for the present,market and future scenario. The cost of heating,c_heat, has a negative impact on LCH. With increase inc_heat, the LCH increases as shown inFig. 6a and b. In themarketand thefuture scenario, the cost of heating contributes as high as 58% to the total expenses. Hence, it is considered as a critical parameter.

FromFig. 6b, the maximum cost of heating was found to be 0.01 and 0.03 V/kWh formarket scenario and future scenario respectively. The intercept on y axis shows the LCH for case

with no cost of heating. These values were found to be approximately 148.5, 2.2 and 0.7V/kg_H₂ for the present, the marketand thefuture scenariorespectively.

Scenario analysis

Out of the three scenarios, LCH inpresent scenariowas found to be the highest, with the cost of membrane being the major contributor.Fig. 7, shows contribution of expenses such as membranes, RED stack without membranes, regeneration system, pumping system, infrastructure and labour to the capital expenses. From Fig. 7, in the present scenario, the membrane and RED stack contribute to approximately 80% of CAPEX, this contribution reduces to more than half in the marketand futurescenario. The labour cost, which includes civil work associated with site preparation, process equipment building, off site services, contributes to 16% of CAPEX in all the scenarios. Pumping system cost increases frompresent scenariotofuture scenariobut has negligible contribution to LCH in all scenarios. The contribution of infrastructure was of low significance (4.4% max. in thefuture scenario).

The maximum limit of cmemto achieve LCH of 3.59V/kg_H₂ formarket scenario(2.86V/m²) is a tenth of cmemin thefuture scenario). Compared to thepresentscenario, the contribution of c_heatto LCH increased from 3% to 50% and 57% formarketand futurescenario respectively. This suggests that as the cost of membrane decreases, and the life of membrane increases, the contribution fromc_heatto the LCH increases (seeTable 7).

In order to improve the accuracy and provide more realistic estimates, the following recommendations are given.

The value of the heat of regeneration,q_th, was taken from a Fig. 4ea. Influence of membrane life time on LCH for AmB REDpresent scenario(left). b. Comparison ofmarketandfuture scenario for membrane lifetime on LCH (right) (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article).

(11)

theoretical study and was not optimized for the present AmB-RED system. Further, to investigate influence of salt type on thermochemical conversion efficiency,h_th, modelling and optimization of regeneration system in a process modelling and simulation software is essential. In the economic model, presented simplified LCH method is used to evaluate the specific cost of hydrogen per kg. This method does not include various operational costs and other capital costs such as consultancy and financing costs. Further, the

economic model is developed for a standard capacity of 1500 kg_H₂/day and hence no upscaling effects were considered. The use of base cost method proposed by Ulrich et al.

[30]. includes these costs and upscaling effect, and thus provides more realistic estimates. The cost of other components in a RED stack were assumed to be the same as the cost of membrane. This may not necessarily be true in case of learning rates. Hence individual component cost estimates are needed. The RED model does not include non-ohmic Fig. 5ea. Influence of inter-membrane distance on LCH forpresent(left) b.marketandfuturescenario (right).

Fig. 6ea. LCH as a function of cost of heating forpresent(left) b.marketandfuturescenario (right).

(12)

resistances, they may influence the hydrogen production rate at low Reynolds number[31].

Conclusions

A simplified LCH model is developed for a thermally driven AmB RED system. The model includes capital costs associated with the regeneration system and operation costs for the waste heat. The scenario analysis includes three different casespresent,marketandfuture; these scenarios differ in the inlet feed solution concentration, membrane properties, residence time, cost of membrane, membrane lifetime. The scenario analysis demonstrates cost of membrane and membrane lifetime as dictating parameters inpresent scenario;

cost of heating and cost of regeneration system, control the levelized cost of hydrogen (LCH) inmarketandfuture scenario.

In case of thepresent scenario membrane replacement cost dominates due to high capital cost of membrane (150V/m²) and limited membrane life of 4 years. While, inmarketand futurescenarios cost of regeneration system influences with contribution to CAPEX as high as 40%. Except in the case of the

present scenario, cost of heating contributes to almost 57% of total expenses, making it a critical parameter that needs to be optimized to further minimize LCH. Parameters such as inter- membrane distance, has a negative effect on the LCH. A decrease in inter-membrane distance increases pumping power but the decrease in heat required to extract salt dominates. The decrease in inter-membrane distance by one fifth of initial value decreases LCH to as low as 1.35V/kg_H₂ in the marketand 0.83V/kg_H₂in thefuturescenario. In future the cost of membrane can reduce to 1.69V/m²with increase in production rate, reduction in production cost, raw material cost, and reduction in membrane thickness. It is clear that AmB RED for hydrogen production has economic potential at membrane cost less than 2.86V/m², and/or, membrane life of 7 years or more. Future efforts must be directed to evaluate performance of potential salts such as lithium bromide that are highly conductive, highly soluble, and may require less heat for recovery. Finally, a thermally driven RED system with membrane cost not exceeding 20V/m²; hydrogen production rate of 3.7 mol/m²/h or higher and cost of heating not more than 0.03V/kWh can make an economically feasible solution for low grade waste heat to hydrogen production.

Fig. 7eComparison of various system elements in relation to CAPEX in the three scenarios.

Table 7eSummary of the results from scenario analysis of the techno-economic analysis.

Present Scenario Market Scenario Future Scenario

LCH 152.97V/kgH2 LCH 3.59V/kgH2 LCH 1.71V/kgH2

Cmem 150V/m² Cmem 2.86V/m² Cmem 3.94V/m²

tmem 4 yrs tmem 7 yrs tmem 10 yrs

_

mH2 0.72 mol/m²/h m_H2 1.19 mol/m/h m_H2 3.71 mol/m/h

Waste heat 480 kWh/kgH2 Waste heat 194 kWh/kgH2 Waste heat 105 kWh/kgH2

Re 0.3 Re 0.13 Re 0.16

CAPEX 43% CAPEX 35% CAPEX 31%

OPEX 10% OPEX 8% OPEX 7%

Cmem;re 44% Cmem;re 7% Cmem;re 5%

Cheat 3% Cheat 50% Cheat 57%

Profit 931,378,389V Profit 0V Profit 11,721,725V

(13)

Acknowledgement

The authors would like to acknowledge Department of Energy and Process Engineering, Norwegian University of Science and Technology, Norway (Project number 70441041) and ENER- SENSE, Norwegian University of Science and Technology, Norway for supporting this research project.

Appendix

r e f e r e n c e s

[1] Quoilin S, Broek MVD, Declaye S, Dewallef P, Lemort V.

Techno-economic survey of organic rankine cycle (ORC) systems. Renew Sustain Energy Rev 2013;22:168e86.https://

doi.org/10.1016/j.rser.2013.01.028.

[2] Miro L, Bru¨ckner S, Cabeza LF. Mapping and discussing Industrial Waste Heat (IWH) potentials for different countries. Renew Sustain Energy Rev 2015;51:847e55.https://

doi.org/10.1016/j.rser.2015.06.035.

[3] Sollesnes G, Helgerud HE. Potensial studie for utnyttelse av spillvarme fra norsk industri. Tech Rep 2009:3e73. Document number 28769.

[4] Forman C, Muritala IK, Pardemann R, Meyer B. Estimating the global waste heat potential. Renew Sustain Energy Rev 2016;57:1568e79.https://doi.org/10.1016/j.rser.2015.12.192.

[5] Kim T, Rahimi M, Logan BE, Gorski CA. Harvesting energy from salinity differences using battery electrodes in a concentration flow cell. Environ Sci Technol

2016;50(17):9791e7.https://doi.org/10.1021/acs.est.6b02554.

[6] Micale G, Cipollina A, Tamburini A. Salinity gradient energy.

Elsevier Ltd.; 2016.https://doi.org/10.1016/B978-0-08-100312- 1.00001-8.

[7] Weiner AM, McGovern RK, Lienhard JH. A new reverse electrodialysis design strategy which significantly reduces the levelized cost of electricity. J Membr Sci 2015;493:605e14.

https://doi.org/10.1016/j.memsci.2015.05.058.

[8] Weiner AM, McGovern RK, Lienhard JH. Increasing the power density and reducing the levelized cost of electricity of a reverse electrodialysis stack through blending.

Desalination 2015;369:140e8.https://doi.org/10.1016/

j.desal.2015.04.031.

[9] Post JW, Goeting CH, Valk J, Goinga S, Veerman J, Hamelers HVM, et al. Towards implementation of reverse electrodialysis for power generation from salinity gradients.

Desalinat Water Treat 2010;16(1e3):182e93.https://doi.org/

10.5004/dwt.2010.1093.

[10] Daniilidis A, Herber R, Vermaas DA. Upscale potential and financial feasibility of a reverse electrodialysis power plant.

Appl Energy 2014;119:257e65.https://doi.org/10.1016/

j.apenergy.2013.12.066.

[11] Turek M, Bandura B. Renewable energy by reverse electrodialysis. Desalination 2007;205(1):67e74.https://

doi.org/10.1016/j.desal.2006.04.041.

[12] Bevacqua M, Tamburini A, Papapetrou M, Cipollina A, Micale G, Piacentino A. Reverse electrodialysis with NH4HCO3-water systems for heat-to-power conversion.

Energy 2017;137:1293e307.https://doi.org/10.1016/

j.energy.2017.07.012.

[13] Veerman J, Vermaas DA. Reverse electrodialysis:

fundamentals. Elsevier Ltd.; 2016.https://doi.org/10.1016/

B978-0-08-100312-1.00004-3.

[14] Burheim OS, Seland F, Pharoah JG, Kjelstrup S. Improved electrode systems for reverse electro-dialysis and electrodialysis. Desalination 2012;285:147e52.https://doi.org/

10.1016/j.desal.2011.09.048.

[15] McCutcheon JR, McGinnis RL, Elimelech M. A novel ammonia-carbon dioxide forward (direct) osmosis

desalination process. Desalination 2005;174(1):1e11.https://

doi.org/10.1016/j.desal.2004.11.002.

[16] McGinnis RL, Elimelech M. Energy requirements of ammonia-carbon dioxide forward osmosis desalination.

Desalination 2007;207(1e3):370e82.https://doi.org/10.1016/

j.desal.2006.08.012.

[17] Luo X, Cao X, Mo Y, Xiao K, Zhang X, Liang P, et al. Power generation by coupling reverse electrodialysis and ammonium bicarbonate: implication for recovery of waste heat. Electrochem Commun 2012;19(1):25e8.https://doi.org/

10.1016/j.elecom.2012.03.004.

[18] Cusick RD, Hatzell M, Zhang F, Logan BE. Minimal RED cell pairs markedly improve electrode kinetics and power production in microbial reverse electrodialysis cells. Environ Sci Technol 2013;47(24):14518e24.https://doi.org/10.1021/

es4037995.

[19] Bandyopadhyay A. Amine versus ammonia absorption of CO2as a measure of reducing GHG emission: a critical analysis. Clean Technol Environ Policy 2011;13(2):269e94.

https://doi.org/10.1007/s10098-010-0299-z.

[20] Nazemi M, Zhang J, Hatzell MC. Harvesting natural salinity gradient energy for hydrogen production through reverse electrodialysis power generation. J Electrochem Energy Convers Stor 2017;14(2):020702.https://doi.org/10.1115/1.4035835.

[21] Schoots K, Ferioli F, Kramer GJ, van der Zwaan BC. Learning curves for hydrogen production technology: an assessment of observed cost reductions. Int J Hydrogen Energy 2008;33(11):2630e45.https://doi.org/10.1016/

j.ijhydene.2008.03.011.

[22] Kwon K, Park BH, Kim DH, Kim D. Parametric study of reverse electrodialysis using ammonium bicarbonate solution for low-grade waste heat recovery. Energy Convers Manag 2015;103:104e10.https://doi.org/10.1016/

j.enconman.2015.06.051.

[23] Naik-Dhungel N. Waste heat to power systems. Tech Rep 2012.https://doi.org/10.1071/EG989133.

[24] Burheim OS. Electrochemical energy storage. In: Engineering energy storage, vol. 1; 2017. p. 1e76.https://doi.org/10.1002/

9781118998151.

[25] Veerman J, Post JW, Saakes M, Metz SJ, Harmsen GJ. Reducing power losses caused by ionic shortcut currents in reverse electrodialysis stacks by a validated model. J Membr Sci Table 8eInput values for parameters described in

equation (20) as a function of inlet solution concentration for stripping column.

Concentration range [M]

a₁ a₂ a₃ a₄ a₅ a₆ a₇

0.025C1¼0.1 12.115 0.26 261.787 0.615 297.46 0.252 0.478 0.1<C10.2 12.836 1.02 258.324 0.612 260 0.165 0.517 0.2<C10.56 13.195 0.686 60.592 0.667 55.934 0.687 0.212 0.56<C12 8.714 0.225 35.796 0.656 45.56 0.758 0.045

(14)

2008;310(1e2):418e30.https://doi.org/10.1016/

j.memsci.2007.11.032.

[26] Zhu X, He W, Logan BE. Influence of solution concentration and salt types on the performance of reverse electrodialysis cells. J Membr Sci 2015;494:154e60.https://doi.org/10.1016/

j.memsci.2015.07.053.

[27] Vermaas DA, Saakes M, Nijmeijer K. Enhanced mixing in the diffusive boundary layer for energy generation in reverse electrodialysis. J Membr Sci 2014;453:312e9.https://doi.org/

10.1016/j.memsci.2013.11.005.

[28] Minke C, Turek T. Economics of vanadium redox flow battery membranes. J Power Sources 2015;286:247e57.https://

doi.org/10.1016/j.jpowsour.2015.03.144.

[29] Kempener F, Ruud, Neumann. Salinity gradient energy:

technology brief. International Renewable Energy Agency (IRENA) (June); 2014. p. 566e71.https://doi.org/10.1109/

OCEANS.1979.1151215.

[30] Fischer UR, Krautz HJ, Wenske M, Tannert D, Kru¨ger P, Ziems C. Hydrogen hybrid power plant in prenzlau, brandenburg, hydrogen science and engineering:

materials, processes, systems and technology 2016;2:1033e51.https://doi.org/10.1002/

9783527674268.ch44(June 2013).

[31] Jalili Z, Wergeland K, Einarsrud KE, Burheim OS. Energy generation and storage by salinity gradient power : a model- based assessment. J Energy Stor 2019.